Document Region Detection for Watermark Embedding Exclusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current digital text watermarking algorithms face challenges in minimizing alteration effects across diverse document contents, such as tables, lists, and math equations, leading to undesirable visual artifacts and reduced information extraction capacity, as they assume uniform text spacing and structures not found in real-world documents.
Innovation Solution
A method using a deep neural network-based text detector to identify and exclude invalid regions from the embedding process, employing a visual representation of documents to predict document invalid regions, including tables and bullet lists, and associating confidence scores with these regions, thereby improving imperceptibility and extraction efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If digital watermarking algorithms process all text regions uniformly, then the embedding process is simple, but visual artifacts appear in tables, lists, and equations
Solution Approach 1:
The patent segments the document into different region types (regular text, tables, lists, equations) and applies different processing strategies to each segment. Invalid regions like tables and equations are excluded from the embedding process, while valid regular text regions are processed. This segmentation resolves the contradiction by maintaining simplicity for valid regions while avoiding artifacts in invalid regions.
Solution Approach 2:
The patent applies different quality standards to different regions of the document. Valid text regions receive the full embedding treatment, while invalid regions (tables, lists, equations) are excluded to maintain their visual quality. This local differentiation resolves the contradiction between uniform processing simplicity and regional visual quality preservation.
2Quantity of substance
If watermarking is applied to all document regions, then the embedding capacity is maximized, but information extraction accuracy decreases due to noise from invalid regions
Solution Approach 1:
The patent extracts and excludes invalid regions (tables, lists, equations) from the embedding process. By removing these regions that would contribute noise to the extraction process, the system maintains high embedding capacity in valid regions while ensuring high extraction accuracy. The exclusion of invalid regions prevents them from acting as noise during extraction.
Solution Approach 2:
The patent performs preliminary detection and classification of document regions before the embedding process. By identifying and marking invalid regions in advance, the system can exclude them from embedding, ensuring that only valid regions contribute to both embedding capacity and extraction accuracy. This preliminary action prevents noise contamination from invalid regions.
3Productivity
If algorithms assume uniform text spacing, then the processing is efficient, but the results are inaccurate for real-world documents with diverse layouts
Solution Approach 1:
The patent employs dynamic region detection that adapts to different document layouts and structures. Instead of assuming uniform spacing, the system dynamically identifies region boundaries and characteristics based on actual document content. This dynamic approach maintains processing efficiency while significantly improving detection accuracy for diverse real-world documents.
Solution Approach 2:
The patent changes the detection parameters and criteria based on the specific document type and layout. Different region types (tables, lists, equations, regular text) are detected using appropriate parameters tailored to their characteristics. This parameter adaptation resolves the contradiction by maintaining efficiency through automated detection while achieving high accuracy for diverse document formats.
Data Source
Figure 1~3(C)
Figure 4~5
Figure 6
AI summary
A method and computer programs for detecting document regions that will be excluded from a watermark embedding process are disclosed. The method comprises converting, by an adapter module, at least one page of a received document into a visual representation thereof, the visual representation keeping the position of the characters of the at least one page; receiving, by a text detector, the visual representation; processing, by the text detector, the visual representation using one or more artificial intelligence algorithms, and returning a list of invalid regions with their associated page positions as a result, wherein each invalid region of the list of invalid regions may have associated thereto a confidence score; and using, by a watermark embedding module or by a watermark extracting module, the list of invalid regions to provide a watermarked document or a message embedded in the document.